splineNetRecon: Network reconstruction based on partial correlation method with shrinkage approach

Description

splineNetRecon reconstructs gene association networks from time-course data. Based on given object of class ExpressionSet, longitudinal data object is created. Subsequantly the function estimates edges using partial correlation method with shrinkage approach applying ggm.estimate.pcor and network.test.edges functions. As a result an object or list of object of class igraph is created.

Usage

splineNetRecon(eSetObject, treatmentType, probesForNR="all", 
            cutoff.ggm=0.8, method="dynamic", saveEdges=FALSE)

Arguments

eSetObject

ExpressionSet object of class ExpressionSet containing log-ratios or log-values of expression for a series of microarrays

treatmentType

a character string containing a type of Treatment defining samples considered for network reconstruction

probesForNR

a vector of character string containing names/IDs used for network reconstruction

cutoff.ggm

number or vector of numbers between 0 and 1 defining cutoff for significant posterior probability; default value is 0.8

method

method used to estimate the partial correlation matrix; available options are "static" and "dynamic" (default) - both are shrinkage methods

saveEdges

if TRUE, .Rdata file with all edges is created; default is FALSE

Value

An object or list of objects of class igraph.
If saveEdges is TRUE, .Rdata file with all possible edges is created.

Details

The input eSetObject must be provided as an object of class ExpressionSet which contains SampleName, Time, Treatment and if applicable Replicates variables (columns) included in the phenotypic data of the eSetObject (pData(eSetObject)). Two types of Treatment defining two groups to compare have to be definied.

Gene association network reconstruction is conducted for a selected type of Treatment. This allows to find regulatory association between genes under a certain condition (treatment). First, a longitudinal data object of the gene expression data with possible repicates is created. This object is used to estimate partial correlation with selected shrinkage method ("dynamic" or "static") with the ggm.estimate.pcor function (for details see ggm.estimate.pcor function help). Finally, the network.test.edges function estimates the probabilities for all possible edges and lists them in descending order (for details see network.test.edges help).

cutoff.ggm can be a single number or a vector of numbers. If more than one value for cutoff.ggm is definied than function returns a list of objects of class igraph for each definied cutoff.ggm value. Otherwise a single object of class igraph with one selected probability is returned.

Examples

Run this code

## load "eSetObject" containing simulated time-course data
data(TCsimData)

## define function parameters 
treatmentType = "T2"
probesForNR = "all"
cutoff.ggm = 0.8
method = "dynamic"

## reconstruct gene association network from time-course data
igr <- splineNetRecon(eSetObject = TCsimData, treatmentType, probesForNR, cutoff.ggm, method)
plot(igr, vertex.label = NA, vertex.size = 3)